Chapter 47

Categories

47.1 Functor
47.1.1 Functor and equivalence
For reference, see [434, 435].
If C and D are categories, a (covariant) functor is a map F : C Ñ D such that
(1) if f : A Ñ B is an arrow in C , then F f : F A Ñ F B is an arrow in D (A and B are objects
in C ),
(2) F pIdA q “ IdF A ,
(3) if g : B Ñ C, then F pg ˝ f q “ F g ˝ F f .
` ˘
A contravariant functor is F : C Ñ D with F pXq P D and F pf q P hom F pY q, F pXq for
every X P C and f P hompX, Y q such that
(1) F pIdX q “ IdF pXq ,
(2) F pg ˝ f q “ F pf q ˝ F pgq.
` ˘ ` ˘
Notice that in the contravariant case, F pf q P hom F pY q, F pXq , and not hom F pXq, F pY q
A functor is an isomorphism if it has an inverse.
When A and B are objects in the category C , we denote by homC pA, Bq the set 1 of arrows
from A to B. Such a set is called an homset. We say that a functor is faithful when its
restriction to each homset is injective. It is full when it is surjective on each homset, i.e. for every
f P hompF A, F Bq, there exists a g P hompA, Bq such that f “ F g. The functor F between C and
D is an equivalence when
(1) it is full,
(2) it is faithful
(3) for every object B in D, there exists an object A in C such that F pAq is isomorphic to B.
The notion of equivalence of categories applies on categories whose objects have a notion of iso-
morphism, such like category of groups, manifold or vector space.

47.2 Direct limit

47.2.1 Direct limit of sets
Let pAi q be a family of set labeled by an ordered set I, and functions fij : Ai Ñ Aj for i § j
such that
— fii is the identity on Ai ,
— fik “ fjk ˝ fij for every i § j § k.

1. In general, it is actually not a set.

2069
2070 CHAPTER 47. CATEGORIES

We say that pI, Ai , fij q is an direct system.

ó The direct limit denoted by A “ limÑ Ai is defined
as the quotient of the disjoint union iPI Ai by the equivalence relation xi „ xj if and only if there
exists a k • i, j such that fik pxi q “ fjk pxj q (xi P Ai and xj P Aj ).
If one sees the maps fij as inclusions, equivalent elements are the ones which finish to be equal.

47.2.2 Direct limit of vector spaces

Consider a diagram V1 Ñ V2 Ñ ¨ ¨ ¨ of vector spaces and linear maps. A direct limit of that
system is a vector space V endowed with a collection of maps Vn Ñ V such that the diagram

Vn / Vn`1

}
V

commutes. Moreover, we impose the following universal property: for every vector space W and
compatible maps
Vn / Vn`1

|
W,
there exists an unique map V Ñ W which is compatible with everything, i.e. such that

Vn

~
V /W

commutes.
One can show that the direct limit exists and is unique in the category of vector spaces. If the
diagram
î is V1 Ä V2 Ä ¨ ¨ ¨ , each of them being a vector subspace of V , then the direct limit is given
by n Vn .
As an example, let consider a sequence of matrix algebras M1 , M2 ,. . . For simplicity, we suppose
that these are full matrix algebras: Mk » nk p q. Put An “ M1 b . . . b Mn ; each of these An
is a matrix algebra and we define An Ñ An`1 ,

T1 b . . . b Tn ބ T1 b . . . b Tn b .

For instance if M1 “ M2 “ 2p q, we have M1 b M2 » 4p q and A1 Ñ A2 ,

ˆ ˙ ˆ ˙
a b a 2 b 2
ބ .
c d c 2 d 2

Notice that the rank is not preserved, for example

¨ ˛
ˆ ˙ 1
1 0 ˚ 1 ‹
fiÑ ˚ ‹.
0 0 ˝ 0 ‚
0

47.2.3 Direct limit in categories

Let now pXi , fij q be a direct system in a category C (the definition is the same as in the case
of sets). The direct limit of this system is an object X of C with arrows Ïj : Xj Ñ X satisfying
Ïi “ Ïj ˝ fij whenever i § j and which fulfils the following universal property: for every object
47.3. CATEGORIES WITH TENSOR PRODUCT 2071

Y P C with arrows Âj : Xj Ñ X such that Âi “ Âj ˝ fij (when i § j), there exists an arrow
u : Y Ñ X which makes the following diagram commutative when i § j

fij
Xi / Xj (47.1)
Ïi Ïj

~
Âi X Âj
u
⇠ ✏ ⌃
Y

All the arrows here are arrows of the category C . One can prove that the direct limit is unique
in every category, but the existence is not always guarantee. In the case of the category of sets
however it exists as we saw an explicit construction as quotient space.

47.3 Categories with tensor product

Définition 47.1 ([436]).
A category C with a functor b : C ˆ C Ñ C and a collection of morphisms

aXY Z : pX b Y q b Z Ñ X b pY b Zq (47.2)

for X, Y, Z P Ob C is a monoidal if
(1) aXY Z is an isomorphism for every X, Y and Z;
(2) for every X, Y, Z, W P Ob C the diagram

` ˘ aXY Z bId ` ˘
pX b Y q b Z b W / X b pY b Zq b W (47.3)
aXbY,Z,W aX,Y bZ,W
✏ ` ✏ ˘
pX b Y q b pZ b W q X b pY b Zq b W

aX,Y,ZbW
*` t Id baY ZW
˘
X b Y b pZ b W q

commutes;
(3) there exists an object I named identity object such that I b I “ I and the functors
X ބ X b I and X ބ I b X are autoequivalences of C .

The commutativity of diagram (47.3) is sometimes called the pentagonal axiom. The trans-
formation a is the associativity constraint.

Exemple 47.2
The category of finite dimensional vector spaces over with the usual tensor product is monoidal
if we set
aXY Z : pX b Y q b Z Ñ X b pY b Zq
(47.4)
px b yq b z ބ x b py b zq,
i.e. the identity map. The unit object is itself seen as one dimensional -vector space. —

A monoidal category is strict if aXY Z “ Id for every objects X, Y and Z and if there exists
an object such that
(1) X b “ b X “ X;
2072 CHAPTER 47. CATEGORIES

(2) the diagram

Id /
X bY pX b q b Y (47.5)
Id aX, ,Y
✏ ✏
X bY /X bp bYq
Id

commutes.
The category of finite dimensional -vector space becomes a strict monoidal category when we
set “ I “ .
An object X in a strict monoidal category is rigid if there exists an object Y and a pair of
morphisms
iX : Ñ X b Y
(47.6)
eX : Y b X Ñ
such that the diagrams
Id
X / bX (47.7)
Id iX bIdX
✏ ✏
Xo X bY bX
IdX beX

and
Id
Y /Y b (47.8)
Id IdY biX
✏ ✏
Y o Y bX bY
eX bIdY

commute. In this case we say that Y is dual of X and we write Y “ X ˚ .

Exemple 47.3
In the case of -vector spaces, the object X ˚ is the usual dual of X. The map eX is the evaluation

eX : X ˚ b X Ñ
(47.9)
– b v fiÑ –pvq

for – P X ˚ and v P X. The map iX is

iX : Ñ X b X˚
ÿ
z fiÑ z ei b e˚i (47.10)
i

where tei u is a basis of X and te˚i u is the dual basis (that∞ is ei˚pej q “ ”ij ). Notice that˚iX is
˚

independent of the choice of the basis. Indeed let us apply i ei bei to the element –bv P X bX.
We have ÿ ÿ ÿ
ei b e˚i p– b vq “ ei p–qe˚i pvq “ –i vi (47.11)
i i i

where
∞ –i and vi are the coordinates of – and v with respect to the basis te˚i u and te˚i u. The number
i –i vi is nothing else than the trace of the map X Ñ X, w fiÑ –pwqv and thus is independent on
the choice of the basis.
These definitions satisfy the axioms of a rigid monoidal category since

v /1bv (47.12)

∞ ˚
✏ ∞ ˚
✏
i ei b ei pvq i ei b ei b v
o
47.3. CATEGORIES WITH TENSOR PRODUCT 2073

∞
commutes because, identifying X with X b ,∞we have i e˚i pvqei “ v. The other diagram is the
same and we found that it commutes because i –pei qe˚i “ –. —

A strict monoidal category is rigid if all the objects are rigid and if for every object X we have
an object Y such that X is isomorphic to Y ˚ .

Exemple 47.4
The category of finite dimensional -vector spaces is rigid since X ˚˚ “ X. —

Let C be a strict monoidal category. A braiding on C is a collection c “ tcX,Y u of functorial

isomorphisms between the functors pX, Y q ބ X b Y and pX, Y q ބ Y b X such that
(1) xX, “ c ,X “ IdX ;
(2) for every objects X, Y and Z the diagrams

pX b Y q b Z (47.13)
cXbY,Z Id
v (
Z b pX b Y q X b pY b Zq
Id IdX bcY,Z
✏ ✏
pZ b Xq b Y pX b Zq b Y

Id ( v cx,Z bIdY
pZ b Xq b Y

and
pX b Y q b Z (47.14)
cX,Y bZ Id
v (
pY b Zq b X pX b Y q b Z
O
Id cX,Y bIdZ
✏
Y b pZ b Xq pY b Xq b Z
h

IdY bcX,Z v Id
Y b pX b Zq
commute.
The commutativity of these two diagrams is sometimes called the hexagonal identities.

Exemple 47.5
The category of finite dimensional -vector spaces is a braided category with

cXY : X b Y Ñ Y b X
(47.15)
x b y ބ y b x.

Let C be a braided strict monoidal category. A balancing on C is a collection b “ tbX u of

automorphisms of the identity functor such that the diagram

bXbY
X bY /X bY (47.16)
bX bby cXY
✏ ✏
X bY o cbY
YXX
/Y bX
2074 CHAPTER 47. CATEGORIES

commutes.
In a balanced category one can define the trace of an arrow f : X Ñ X. The trace is the arrow

TrX pf q : Ñ (47.17)

given by
b bIdX ˚
/ X b X ˚ f bId / X b X ˚ X
e
iX
/ X b X ˚ X˚ / . (47.18)

Exemple 47.6
In the case of the finite dimensional -vector spaces, we have the balancing bX “ IdX . We check
that the trace of a linear map f : X Ñ X is the well known map:
∞ ∞ ∞ ` ˘ ∞
1 /
i ei b e˚i /
i f pei q b e˚i / ˚
i ei f pei q “ i f pei qi . (47.19)

Définition 47.7.
A rigid balanced braided monoidal category is a ribbon category.